AI summary by TwinMind
How AI and Democracy Can Fix Each Other | Divya Siddarth | TED
TED
11m 2s
Personal Journey and Philosophy
- Speaker started with belief in science power - modifying DNA in kitchen at age 12
- Studied computational biology at Stanford, converted to technology belief
- Observed most impactful technologies built with/for collective, not individual profit
- Contrasted billions pumped into addiction-fueling social apps vs. internet built with public resources on open standards
- Led to democracy focus: technology expands capabilities, democracy decides what to do with capability
Collective Intelligence Project Formation
- Co-founded with Saffron Hang after leaving Google DeepMind and Microsoft
- Named "Collective Intelligence Project" - democracy as early form of collective intelligence
- Mission: building new democratic governance models for transformative tech
- Approach: decentralized input from diverse sources producing decisions better than sum of parts
AI Governance Research Findings
- Worked in India, US, UK, Taiwan with policymakers and companies (OpenAI, Anthropic)
- Key finding: people willing/able to have complex conversations on nuanced topics
- AI risk concerns survey: top concern was "over-reliance on systems we don't understand" (out of 100+ risks)
- Interviewed diverse participants (veteran in midwest, young teacher in South)
- Minimal polarization observed: few divisive statements vs. hundreds of consensus statements
- Specific consensus statistics:
- 75% agree AI should protect free speech
- 90% agree AI should not be racist/sexist
- Only 50% think AI should be funny
Collective Constitution Project with Anthropic
- Retrained world's most powerful language model on principles written by 1,000 representative Americans
- Participants: ordinary people, not AI developers/regulators/elite university researchers
- Process: co-writing constitution for AI training vs. researcher-developed model
- Results: people's model equally capable and more fair than researcher model
- Conclusion: people with little AI experience outperformed full-time researchers in building fairer chatbot
- Participant quote: "They may be experts in AI but I have 8 grandchildren. I know how to teach values"
Taiwan Collaboration
- Working with Minister Audrey Tang and Ministry of Digital Affairs
- Process: asking Taiwan's millions of citizens about desired AI future
- Using input for both legislation and building
- Key outcome: public excitement for "public option for AI"
- Built on shared public data
- Reliably safe
- Allows communities to access, benefit from, and adjust to their needs
Three-Way Framework: Progress, Safety, Democracy
- Rejected false trade-offs: don't sacrifice democracy for technological progress
- Argued against concentration of power for safety from risks
- Thesis: impossible to have progress, safety, OR democratic participation without all three
- Warning: settling for only two leads to centralized control or chaos
- Evidence: each project advanced all three simultaneously through:
- Cutting-edge democratic AI models
- Public expertise for understanding diffuse risks
- Co-ownership models for digital commons
Future Vision and Possibilities
- Current collective intelligence systems far from optimal potential
- Hypothetical improvements if starting over:
- Better separation of financial from political power
- Thousands of new corporation/bureaucracy models
- Voices of natural elements/future generations built in
- Observation: new technologies create paradigm shifts enabling new collective intelligence systems
- Opportunity: use brief openings for change to create new ways of living well together
Conclusion
- Quote from Indian author Arundhati Roy: "Another world is not only possible, she is on her way. On a quiet day I can hear her breathing"
- Vision: shift systems toward using democracy as solution to build desired worlds
- Call to action: "The future is up to us. We have a world to win"
